🚨 New working paper 🚨
Demand for LLMs: Descriptive Evidence on
Substitution, Market Expansion, and Multi-Homing

A key question for the business of AI is the extent to which LLMs are differentiated from each other. I use data from OpenRouter to take a first look.

andreyfradkin.com/assets/deman...

I got an email yesterday afternoon that my NSF SPRF Postdoctoral Fellowship was terminated. My grant focused on testing interventions to address online misinformation and I was 8 months into a two year appointment.

Thread: Obviously I'm very pleased at this recognition, but I'm sharing this for other reasons. 1st, many have the impression that people who have been successful professionally have walked a straight path - got the right degree, right 1st job, succeeded right away.

Forthcoming in AEJ: Applied Economics: "Labor Supply and Entertainment Innovations: Evidence From the U.S. TV Rollout" by George Fenton and Felix Koenig. www.aeaweb.org/articles?id=...

Here’s a piece I wrote with @rfisman.bsky.social in the NYT, collecting some of our thoughts about what the oligarchic embrace of Trump can do to our economy and our politics. www.nytimes.com/2025/01/21/o...

How do great research collaborations happen? 🌟
Learn how Dr. Katy Milkman's (@katymilkman.bsky.social) partnership with CVS Pharmacy unfolded and what tips she has for partnering with major organizations. chibe.upenn.edu/blog/how-to-...

#QJE Feb 2025, #9, “The Long-Run Impacts of Public Industrial Investment on Local Development and Economic Mobility: Evidence from World War II,” by Garin (@andygarin.bsky.social) and Rothbaum: doi.org/10.1093/qje/...

A List of Predictions Made in 1925 About 2025

🧵

Thank you, Kristina!

Takeaway: A key trade-off for policymakers (e.g., DMA) and managers to keep in mind: Quick wins with external data vs. building self-sufficient internal capabilities. ⚖️

Deep gratitude to the Management Science team, reviewers, & everyone who offered feedback. 🙏

(6/6)

2️⃣ The trade-off: Over time, the negative impact of removing the external API halved. Why? Relying on external data limits internal data accumulation—critical for long-term product improvement. (5/)

Key Findings: 🧑‍🔬

1️⃣ The value: Removing external data access by removing API input led to ↓ a search-suggestion CTR of 4.6%. This drop has downstream impact: A 10% ↓ in suggestion CTR → 1.85% ↓ in CTR on top search results on the main search page. (4/)

The study involved 2M+ users over 108 days. Half of them saw autocomplete search suggestions from both the leading external API + our partner firm. The other half? Just our partner firm’s autocomplete suggestions (i.e., no external data through API). (3/)

We explore how data sharing (through a leading search engine’s autocomplete API) affects smaller players’ product performance. 📊

We conducted a large-scale field experiment with a tech firm in the search market. (2/)

My first post on Bluesky! Excited to share my paper with Yixing Chen and Xiaoxia Lei "Trade-offs in Leveraging External Data Capabilities: Evidence from a Field Experiment in an Online Search Market" has been accepted for publication in Management Science! papers.ssrn.com/sol3/papers....

🚨In Science🚨
Conspiracy beliefs famously resist correction, ya?
WRONG: We show brief convos w GPT4 reduce conspiracy beliefs by ~20%!
-Lasts over 2mo
-Works on entrenched beliefs
-Tailored AI response rebuts specific evidence offered by believers
www.science.org/doi/10.1126/...
1/
[X->BSky repost]

New publication (4+ years in the making): “The Diffusion and Reach of (Mis)Information on Facebook”. shorturl.at/VE2fU
We analyze the propagation of 1B+ posts across content moderation regimes, with @davidlazer.bsky.social @jatucker.bsky.social @taliastroud.bsky.social @annenbergpenn.bsky.social

1/ A thread on the madness of Dem fundraising and why it needs to change 🧵

📈 Look at this chart of political fundraising emails I received in 2024:
- 3,212 fundraising emails in total.
- 48 in a single day!

Badgering your supporters is disrespectful, annoying, and, frankly, counterproductive.

``Consumer surplus lost after a ban on cash payments is large, equal to 40-50% of total expenditures on cash-fare Uber rides. Welfare losses mostly fall on the least-advantaged households."

Recently accepted to REStud, from Alvarez and Argente:

www.restud.com/consumer-sur...

Recently accepted to #REStud, ``The Long-Run Labor Market Effects of the Canada-U.S. Free Trade Agreement,'' from Kovak and Morrow:

www.restud.com/the-long-run...

Next Tuesday, 14h CET, Yaron Yehezkel (Tel-Aviv Un.) will present "Competing for Cookies: Platforms’ Business Models in Data Markets with Network Effects" (with S. Markovich)
Discussant: Tat-How Teh (Nanyang Technological Un.)
at the TSE Economics of Platforms seminar.
www.tse-fr.eu/online-semin...

this guy on youtube had both his son and daughter do the trolley problem (at the same age)

Please come work with me at the Stanford Digital Economy Lab!

We're seeking a postdoctoral associate to explore exciting work focused on transformative AI.

Accepting applications now. The link is in the next post.